A support vector method for multivariate performance measures

By Thorsten Joachims. 2005

930 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Optimization & Training, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Optimization & Training

In Signal Processing

Citations in the corpus

Painless Stochastic Gradient: Interpolation, Line-Search, and Convergence Rates (2019, background): “These problems encompass both convex minimization and saddle point problems arising in robust supervised learning [8, 84] and learning with non-separable losses or regularizers [4, 34].”
Search-based structured prediction (2009, neutral): “These difculties have been partially dealt with by successive algorithms (Tsochantaridis et al., 2005; Taskar et al., 2005; Joachims, 2005 ).”
Recommendations as Treatments: Debiasing Learning and Evaluation (2016, background): “Solving this training objective for other δ u,i ( Y, ˆ Y ) that are non-differentiable is more challenging, but possible avenues exist (Joachims, 2005; Chapelle & Wu, 2010).”
Satisfying Real-world Goals with Dataset Constraints (2016, methodology): “Joachims [16] indirectly optimizes upper-bounds on functions of sp(D), sp(D), sn(D), sn(D) using a hinge loss approximation.”
Kernel methods in machine learning (2007, neutral): “Other sparse vector kernels are also conceivable, such as one that maps a text to the set of pairs of words that are in the same sentence (Joachims [77] and Watkins [151]). n-grams and suffix trees.”
APRIL: Active Preference-learning based Reinforcement Learning (2012, neutral): “where weight vector wt is obtained by standard preference learning, solving quadratic constrained optimization problem P [28,15]:”
Stochastic Variance Reduction Methods for Saddle-Point Problems (2016, methodology): “Non-separable losses may occur when (a) predicting by affine functions of the input s and not penalizing the constant terms (in this case defining the loss functio s as the minimum over the constant term, which becomes non-separable) or (b) using structured outpu t prediction methods that lead to co”
Sentence Compression Beyond Word Deletion (2008, methodology): “The model is trained using SVM struct , a large margin method for structured output problems (Joachims, 2005; Tsochantaridis et al., 2005).”
The K-armed Dueling Bandits Problem (2012, methodology): “We will present algorithms which achieve identical regret bounds for both formulations (up to constant factors) by assuming a property called stochastic triangle inequality , which is described in the next section.”
Unbiased Metric Learning: On the Utilization of Multiple Datasets and Web Images for Softening Bias (2013, methodology): “ψ(q, y) is the partial order feature [11] with the following form:”
Hinge Rank Loss and the Area Under the ROC Curve (2007, methodology): “…the hinge rank loss by the (standard) hinge loss cannot be proven theoretically, it is strongly supported not only by our experiments in Section 7 but also by many experimental ﬁndings in the literature, where standard SVM training yielded surprisingly high AUC values [3,4,5,6], cf. also Section 8.”
From region similarity to category discovery (2011, methodology): “Metric learning to rank (MLR, Algorithm 1) [14] is a metric learning extension of the Structural SVM algorithm for optimizing ranking losses [10, 21].”
Optimizing F-measure: A Tale of Two Approaches (2012, methodology): “Joachims (2005) gave an eﬃcient algorithm for maximizing a convex lower bound of F-measures for support vector machines, and showed it worked well on text classiﬁcation.”
Optimizing Intersection-Over-Union in Deep Neural Networks for Image Segmentation (2016, methodology): “For example, Joachims [1] proposed a multi-variate SVM formulation for optimizing a range of nonlinear performance measures including F1-Score and ROC-area.”
Scalable Learning of Non-Decomposable Objectives (2016, methodology): “Several recent works, starting with the seminal work of Joachims [11] have addressed the challenge of optimizing various rank-based objectives.”